r""" Compute the quantization error if we use start_bin to end_bin as the min and max to do the quantization.

Copy node from the original graph and replace the input names.

Placeholder|px ──────► Abs|y1 ─────────────────┐ │ . │ │ .ctl_dep

------------------ px -> Mul -> | -> Add -> Mul -> | -> Relu -> Sqrt / | / / | py - c ->

px -> Add --> Mul -> Relu -> ABS -> Add -> Sqrt / \_/ / \ # noqa: W605 py -

Check whether a view-kind node should be considered as an inplace node. let A defines a node whose kind is defined in self._view_kin

This method is stateful, it cache safe inplace ops flags

r"""Non-linear parameter search. An approximation for L2 error minimization for selecting min/max. By selecting new min/max, we filter

Turn 'foo.bar.name' into ['foo.bar', 'name']

Add scalar type and dim to plain TensorType.

Add input tensors to the subgraph.

Add output nodes to this subgraph.